Common Misconceptions about the Consumer Price Index

The BLS responds to criticism about the CPI. The answers are on the continuation page:

Common Misconceptions about the Consumer Price Index: Questions and Answers, BLS: An August 2008 Monthly Labor Review article by BLS economists John Greenlees and Robert McClelland reviews and analyzes some common misconceptions about the Consumer Price Index (CPI.) Those analyses are summarized here:

1. Has the BLS removed food or energy prices in its official measure of inflation?
2. The CPI used to include the value of a house in calculating inflation and now they use an estimate of what each house would rent for -- doesn't this switch simply lower the official inflation rate?
3. When the cost of food rises, does the CPI assume that consumers switch to less expensive and less desired foods, such as substituting hamburger for steak?
4. Is the use of "hedonic quality adjustment" in the CPI simply a way of lowering the inflation rate?
5. Has the BLS selected the methodological changes to the CPI over the last 30 years with the intent of lowering the reported rate of inflation?
6. Does the Bureau of Labor Statistics calculate the CPI the same way as other nations? Do any differences in method keep the US CPI lower than the CPIs of those other nations?

Has the BLS removed food or energy prices in its official measure of inflation?

No. The BLS publishes thousands of CPI indexes each month, including the headline All Items CPI for All Urban Consumers (CPI-U) and the CPI-U for All Items Less Food and Energy. The latter series, widely referred to as the "core" CPI, is closely watched by many economic analysts and policymakers under the belief that food and energy prices are volatile and are subject to price shocks that cannot be damped through monetary policy. However, all consumer goods and services, including food and energy, are represented in the headline CPI.

Most importantly, none of the prominent legislated uses of the CPI excludes food and energy. Social security and federal retirement benefits are updated each year for inflation by the All Items CPI for Urban Wage Earners and Clerical Workers (CPI-W). Individual income tax parameters and Treasury Inflation-Protected Securities (TIPS) returns are based on the All Items CPI-U.

The CPI used to include the value of a house in calculating inflation and now they use an estimate of what each house would rent for -- doesn't this switch simply lower the official inflation rate?

No. Until 1983, the CPI measure of homeowner cost was based largely on house prices. The long-recognized flaw of that approach was that owner-occupied housing combines both consumption and investment elements, and the CPI is designed to exclude investment items. The approach now used in the CPI, called rental equivalence, measures the value of shelter to owner-occupants as the amount they forgo by not renting out their homes.

The rental equivalence approach is grounded in economic theory, receives broad support from academic economists and each of the prominent panels, and agencies that have reviewed the CPI, and is the most commonly used method by countries in the Organization for Economic Cooperation and Development (OECD). Critics often assume that the BLS adopted rental equivalence in order to lower the measured rate of inflation. It is certainly true that an index based on home prices would be more volatile, and might move differently from other CPI indexes over any given time period. However, when it was first introduced, rental equivalence actually increased the rate of change of the CPI shelter index, and in the long run there is no evidence that the CPI method yields lower inflation rates than some other alternatives. For example, according to the National Association of Realtors, between 1983 and 2007 the monthly principal and interest payment required to purchase a median-priced existing home in the United States rose by 79 percent, much less than the rental equivalence increase of 140 percent over that same period.

When the cost of food rises, does the CPI assume that consumers switch to less desired foods, such as substituting hamburger for steak?

No. In January 1999, the BLS began using a geometric mean formula in the CPI that reflects the fact that consumers shift their purchases toward products that have fallen in relative price. Some critics charge that by reflecting consumer substitution the BLS is subtracting from the CPI a certain amount of inflation that consumers can "live with" by reducing their standard of living. This is incorrect: the CPI's objective is to calculate the change in the amount consumers need to spend to maintain a constant level of satisfaction.

Specifically, in constructing the "headline" CPI-U and CPI-W, the BLS is not assuming that consumers substitute hamburgers for steak. Substitution is only assumed to occur within basic CPI index categories, such among types of ground beef in Chicago. Hamburger and steak are in different CPI item categories, so no substitution between them is built into the CPI-U or CPI-W.

Furthermore, the CPI doesn't implicitly assume that consumers always substitute toward the less desirable good. Within the beef steaks item category, for example, the assumption is that consumers on average would move up from flank steak to filet mignon if the price of flank steak rose by a greater amount (or fell by less) than filet mignon prices. If both types of beef steak rose in price by the same amount, the geometric mean would assume no substitution.

In using the geometric mean the BLS is following a recognized best practice for statistical agencies. The formula is widely used by statistical agencies around the world and is recommended by, for example, the International Monetary Fund and the Statistical Office of the European Communities.

Is the use of "hedonic quality adjustment" in the CPI simply a way of lowering the inflation rate?

No. The International Labour Office refers to the hedonic approach as "powerful, objective and scientific". Hedonic modeling is just one of many methods that the BLS uses to determine what portion of a price difference is viewed by consumers as reflecting quality differences. It refers to a statistical procedure in which the market valuation of a feature is estimated by comparing the prices of items with and without that feature. Then, for example, if a television in the CPI is replaced by one with a larger screen and higher price, the BLS can make an adjustment to the price difference by estimating what the old television would have cost had it had the larger screen size.

Many of the challenges in producing a CPI arise because the number and types of goods and services found in the market are constantly changing. If the CPI tried to maintain a fixed sample of products, that sample quickly would shrink and become unrepresentative of what consumers were purchasing. Each time that an item in the CPI sample permanently disappears from the shelves, the BLS has to choose another, and then has to make some determination about the relative qualities of the old and replacement item. If it did not-for example, if it treated all new items as identical to those they replaced-significant upward or downward CPI biases would result.

Critics often incorrectly assume that BLS only adjusts for quality increases, not for decreases, and that hedonic adjustments have a large downward impact on the CPI. On the contrary, BLS has used hedonic models in the CPI shelter and apparel components for roughly two decades, and on average hedonic adjustments usually increase the rate of change of those indexes. Since 1998, hedonic models have been introduced in several other components, mostly consumer durables such as personal computers and televisions, but these newer areas have a combined weight of only about one percent in the CPI. A recent article by BLS economists estimated that the hedonic models currently used in the CPI outside of the shelter and apparel areas have increased the annual rate of change of the All Items CPI, but by only about 0.005 percent per year.

Has the BLS selected the methodological changes to the CPI over the last 30 years with the intent of lowering the reported rate of inflation?

No. The improvements chosen by the BLS that some critics construe to be a response to short term political pressure were, in fact, the result of analysis and recommendations made over a period of decades, and those changes are consistent with international standards for statistics. The methods continue to be reviewed by outside commissions and advisory panels, and they are widely used by statistical agencies of other nations.

Moreover, the sizes and effects of the changes implemented by the BLS are often over-estimated by critics. Some have argued that if the CPI were computed using the methods in place in the late 1970s, the index would now be growing at a rates as high as 11 or 12 percent per year. Those estimates are based on the belief that the use of a geometric mean index lowered the annual rate of change of the CPI by three percentage points per year, and a belief that other BLS changes, such as the use of hedonic models and rental equivalence, have lowered the growth rate of the CPI by four percentage points per year.

Neither belief is supported by evidence. BLS calculations have shown that the geometric mean formula has reduced the annual growth rate of the CPI by less than 0.3 percentage points. Hedonic quality adjustments for shelter regularly increase the rate of change of the CPI, and those for apparel have had both upward and downward impacts at different points in time and for different types of clothing. The BLS estimates that the overall impact of hedonic quality adjustments in use in other categories has been extremely small. Furthermore, if the CPI were using the pre-1983 asset-based method instead of rental equivalence to measure homeowner shelter cost it would yield a sharply lower current measure of shelter inflation, given that house prices are now declining in many parts of the country.

Does the Bureau of Labor Statistics calculate the CPI the same way as other nations? Do any differences in method keep the US CPI lower than the CPIs of those other nations?

Yes, the methods described above are used widely by nations in the OECD and the European Union. A recent report shows that rental equivalence is the most common method used to measure changes in the cost of shelter by the OECD - with 13 of 30 nations employing it. The next most common method is for a nation to omit shelter from the CPI. The hedonic method of quality adjustment is used by at least 11 of the 29 other OECD nations, and five of the G-7 nations. Eurostat reports that the geometric mean is used by 20 of 30 countries for its Harmonized Indices of Consumer Prices.

Each nation's inflation experience is the result of its unique economic circumstances, so comparing the change in the U.S. CPI-U with inflation rates in other countries does not gauge the accuracy of U.S. inflation measures. Nevertheless, over the 1997-2007 period the U.S. CPI-U increased faster than the CPIs of 16 of the other 29 OECD nations, and faster than the CPIs of all of the other G-7 nations, including Canada, the United States' largest trading partner. Similarly, between the first quarters of 2007 and 2008 the U.S. CPI rose by more than the CPIs of 20 of the other 29 OECD nations and by more than any of the other G-7 nations, including Canada.

Find out more in "Addressing misconceptions about the Consumer Price Index" in the August 2008 Monthly Labor Review.

Update: Also see Jim Hamilton's "Shadowstats debunked."

Comments

[cm]

Apparently the time has come to explain to the "whiners" that they have "misconceptions".

[esb]

No, they have misunderstandments.

Or is it misunderstandmentations?

[Eric]

The BLS is using NAR data as justification? Not exactly confidence inspiring.

[esb]

Questions 4 and 5 are just too, too amusing.

Oh, and they left out question 7,

Does the BLS recruit exclusively from graduates of the
CMI (Con Men's Institute).

It was left out due to the policically incorrect nature of the word, "Men's."

[spencer]

esb-- do you usually try to influence people by demonstrating how ignorant you are, or is this an exception?

[Bruce Wilder]

I don't find the condescension persuasive, or the presentation generally fair-minded.

The brief reference to how hedonics are used, was frightening. Maybe, that's because I actually know something about hedonic price indices, having worked on one for automobiles, back in the day. Properly done, for a product like an auto, with a lot of characteristics, which can dominate nominal price, an hedonic adjustment can be very revealing. But, you can easily get lost, confusing a consumer decision to buy "more" of a product in the form of a higher quality product, with a change in the price of the product. You cannot possibly do this properly, while maintaining strictly within the bounds of fixed-weight market basket, but if you also allow price-substitution trade-offs to redefine the market basket, overdetermination means that you are simply projecting your own judgments into the index. The arbitrariness necessary to the integrity of the index is lost.

Ultimately, price indexes are always wrong, being linear approximations of non-linear trends, but their integrity and objective character depends upon their arbitrary construction, by rule, and not by subjective judgment. The use of various forms of the CPI and other indexes as instruments of policy may have over-strained that integrity.

[Mark Thoma]

Hedonic adjustments are covered on pages 7-10 of the more extensive document linked above. The Q and A is just a summary.

[cm]

Bruce Wilder: It is also easy to confuse the fact that "last year's model" is no longer for sale with "consumer preference".

As for computers, I have yet to be convinced that this year's edition loaded up with this year's software plus the ever more expanded need for virus checkers, firewall filters, resource-hungry flashy interfaces that don't add new functionality and confound users with rearranged controls etc., are really enhancing end user productivity.

At work we got "Office 2007" installed which is considerably slower than the previous version, on the same machines.

[Roger Chittum]

Assuming some conversational linkage from immediately prior posts to this one, is Mark representing that the CPI methodology has always been "correct" from time to time over all of the last 60 years, that methodological changes were only made as real world changes required, and that, for example, there were no economically significant hedonic changes before the date the BLS began to adjust for hedonic changes? I ask because, really, the question was whether the knuckles in the income graphs are "real" or mostly due to BLS "improving" its methodology. I come away with the impression that the "flawed" post-war methodologies may have overstated inflation and that these problems got fixed over time, and that if the "more correct" current methods had been applied from 1947 to 1973 the income growth rates in that era would have been even higher than reported in Mark's previous post. Thus, it seems, the graphs are more likely to understate than to overstate the drama of what began happening circa 1974.

[SanFranciscoJim]

You know, non-kooks like the editors at The Economist have argued that inflation, properly stated, is 2-3% higher than what the BLS current puts it at. Though this may not be true anymore, in the era of falling home prices. I can dig up the article if anyone is interested.

I think if you are upper-middle class, and your biggest expenditures are mortgage, tuition and health care, and you shop at Whole Foods, inflation seems understated and has been for a while. But if you buy lots of good from China, it might even be understated, though perhaps not this year.

[marcello]

This to me is the money quote:

>the CPI's objective is to calculate the change in the
>amount consumers need to spend to maintain a constant
>level of satisfaction.

WTF? CPI is ALWAYS represented as the price change in a fixed _basket of goods_, not a _fixed consumer satisfaction level_. You know, a measure of how much more/less our paycheques can buy now than before - a useful bit of information, to say the least.

How the f*** can one believe even an order-of-magnitude guesstimation of a ridiculously subjective notion as "constant level of satisfaction", much less a result good to a couple of decimal places ? That's utterly preposterous, and it matters not a rat's behind that "economic theory" or this or that international organization does the same thing. So if suddenly the country prefers hot dogs to steak, the CPI drops ??? they must be joking.

And sure, they would have you believe that it is all just a coincidence that CPI is vastly smaller now than before the changes, as ShadowStats has shown. The unfiltered gall.

It's true: economics does make astrology look respectable

[For sale]

Somebody doesn't understand that this debunks Shadowstats...

Kind of funny. Especially how poorly you understand the calculations, yet you rail away anyhow making a fool of yourself.

Ha - the left has plenty of wingnuts, especially where the Fed or any kind of government stats is concerned.

The unfiltered stupidity.

I should be selling tin foil hats.

[Common Sense Test]

A lot of words, but the CPI doesn't pass the common sense test. A generation ago, a single median wage could easily support a family. Today, the CPI adjusted median wage is higher, but a single median wage can no longer reasonably support a family.

Something is very wrong.

[roger]

I admire the rhetorical structure of that defense. It reminds me of the rhetoric I´ve just heard at the Republican convention. Begin with a ¨misunderstanding of the food and fuel component that is really just ignorance. It is as if I said, many people dont believe in astrology. Among the common misconceptions: astrologers believe the world is flat. No. Professional astrologers agree the world is round. It has always been round, and as we improve our methodology, we take into account its roundness.

So sad. Establishment economics, with its deep ties to a center right worldview, and its bizarre notion that it is a science like physics, except even better, Freakonomically explaining everything in terms of the model of rational choice, is heading towards a crisis of confidence. You cant keep making shit up.

[marcello]

BTW, John Williams of ShadowStats has been informed of the BLS article, says there is nothing new at all in it and he has answered those issues many times, yet he may do so again anyway, I guess to debunk the debunkers.

[freelunch]

Every step in the process that the BLS uses is defensible and necessary, but, as noted above, these are approximations. Hedonics will always have critics, no matter how determined and, as we see from model-year-end clearance sales, you could come up with a very large value for the improvement from one year to the next, though most people realize that most of the pricing difference has to do with resale value and corporate marketing decisions, not improvement in the product.

In a perfect world, the inflation rate would be measured monthly, based on the difference between what it cost to buy last month's goods and services last month and what it would have cost this month to buy those same goods and services, even though there will be a small change in the basket for the month. Rebalancing the basket to reflect the entire economy each month keeps the inflation indicator current, but relies on information that won't be available for months.

Eventually, we will have better data and critics like ShadowStats and the Economist will help us get to that better data more quickly.

[rdaneel]

The many comments attacking the BLS, and praising Shadowstats, at this site, and at Jim Hamilton's site, are depressing to me as an economist. The BLS article is an extremely convincing refutation of recent criticisms. Several points that some of these critics might want to take into account are the following:

1. As BLS states, with numerical examples, the changes made to CPI measurement methods have not in general had large effects on the average measured inflation rate.
2. Shadowstats does not actually recompute the CPI using old methods. Rather, it assumes some large fixed deviation between inflation measured using old methods and inflation measured using new methods. This fixed adjustment makes no sense, because the various new methods will have different effects on reported inflation in different time periods.
3. Right now, with dramatic declines in the asset prices of owner occupied housing, there is a strong possibility that inflation calculated using the old methods would be LOWER than inflation as calculated now by the BLS. Therefore, if you for some reason believe that measured inflation is below "true inflation", going back to the old measurement methods will not help.

[Half As Much is Not the Same]

It is really not reasonable that the same CPI adjusted amount of money should buy half as much living standard as it did in the past. No matter how smooth the words, few will believe something that conflicts with common sense.

[roger]

Right now, with dramatic declines in the asset prices of owner occupied housing, there is a strong possibility that inflation calculated using the old methods would be LOWER than inflation as calculated now by the BLS.´´

This is certainly true. Just as it is true that the housing price spike was certainly not accurately recorded by the BLS. If, of course, an accurate reflection of the housing market prices were recorded by the BLS over the 2000-2006 period, the Fed simply couldn´t have run the easy money policy that it did. So there are consequences. Now, of course, there´s a big divergence between falling house prices and rising institutional prices - for education, health, etc. - and consumer product prices. These divergences ought to show economists their instruments are broken, and that the lack of good instruments has messed up the whole fiscal machine. It should have occured to the Fed that if the average household was taking out massive equity loans as the equivalent for the wage increases that the Fed had successfully crushed in the service of big business, that this would have an odd inflationary effect - such that it is possible that increasing the interest rate could lead to real wage increase demands, as those interest rates will take a major bite out of the average household in ways undreamt of in 1980.

But why worry when you can circle the horses? When the models said that the sun circled the earth, and experts from Ptolemy on could be referenced if there was any conspiracist type of question by the plebians out there, at least a few scientists started questioning the model.

But no. Economists have been consistently over optimistic this year. Wonder why? Could it be that their denial of motives except the highest one of the truth (which is believed by nobody), the condescending tone of their defenses, and their belief that an argument consists of repeating what you have already said, once, is sufficient. Who will deliver us from the children of the narrowing and triumphalism of the current state of economics?

[Blissex]

«Rebalancing the basket to reflect the entire economy each month keeps the inflation indicator current, but relies on information that won't be available for months.»

Every month is too frequent. Countries with some integrity in their inflation stats change the basket every about 10-20 years.

The BLS instead redefined the basket every time it wants to keep the numbers down, assuming that every basket it comes up with has the same "level of satisfaction":

«the CPI's objective is to calculate the change in the amount consumers need to spend to maintain a constant level of satisfaction.»

Buffoons.

[Bob]

It is amazing to see so much talk of inflation without a single mention of using money supply to calculate it. Prices are determined almost entirely by the amount of production and the money supply - supply of real goods and demand of monetary units. Price inflation should reflect economic growth or recession and monetary inflation or deflation. The problem here is that there is no means to measure production separate from monetary terms, as production must take into account the value of that produced, which can either be a subjective estimation or measured in monetary units.

Thus, the only truly objective measure of inflation is measuring changes to the money supply. This gives a more accurate measure of the change in value of a monetary unit for a given unit of time as well, as production changes are not fundamentally related to the change in monetary stock.

Consumer prices are generally consequential, not speculative of changes in the money supply. Thus, changes in the money supply will take time to be reflected in consumer prices.

ShadowStats is hogwash. BLS CPI + 7% is more subjective than hedonics or satisfaction levels, yet requires much less work.

The truth is that any price index is going to be inaccurate, incomplete, and probably subjective as well.

I'd say the BLS is wrong, based upon the common sense evaluation - lots more dual-income households yet not a terribly higher standard of living. Of course, we average 6-8% growth in money supply and 2-3% economic growth. We should expect 3-5% annual price inflation. Given that the BLS's CPI is consistently lower than these numbers, I would assume it is wrong; but not a blatant lie like the cranks will proclaim.

[Bob]

I meant 3-6% annual price inflation. BLS CPI hasn't hit a rate over 3.5% since the early 90's. It is too conservative, but it's better than junk science.

Also, if I didn't imply it, with constant money supply and increasing production, we should see price deflation, which would indicate the true cost (in human labor or invested capital) of most goods would decrease over time.

BTW, I am using Mises.org 's True Money Supply (TMS) stat as my measurement of money supply. Most of M3 doesn't measure what many would define as money, and much of M2 is also not money. MZM even includes some items that TMS does not. Despite appearing so conservative, the growth rates are what really matters - 6-8% annualized growth is a heavy amount of inflation.